A Topological Data Analysis(TDA) Based Lightweighting and Quality Analysis of Large-Scale 3D Point Cloud for Ship Blocks
Joonsoo Kim, Ki-Seok Jung, Dong-Kun Lee · Journal of the Society of Naval Architects of Korea · 2025
The adoption of various Industry 4.0 technologies in the shipbuilding sector has recently accelerated, with particular attention being given to techniques utilizing three-dimensional(3D) point cloud data acquired from LiDAR scanners. However, the raw 3D point cloud data presents significant challenges in storage, processing, and analysis due to its high-dimensional and large-scale nature. This necessitates a lightweighting process following initial post-processing. Accordingly, this study proposes a lightweighting technique based on Topological Data Analysis(TDA) for the efficient handling of 3D point clouds. TDA is a methodology that utilizes persistent homology, an adaptation of the core topological concept of homology tailored for data analysis. Based on TDA, feature points were extracted by considering the size and complexity of the point cloud, guided by the lifetime of homological features in the persistence diagram, which plots their birth and death. Specifically, features with longer lifetimes are considered topologically significant; by selectively extracting these persistent points, the overall size of the dataset is reduced. For field validation, the TDA-based lightweighting method was applied to actual measurement data from ship blocks. The results demonstrate that the proposed technique excels at preserving geometric features while significantly reducing data volume. Consequently, the lightweighting technique developed in this research is anticipated to enhance the utility of datasets for deep learning applications in the future.